Re: Oh, well, looks like atoms are not possible

[Chris Marrin <chris@...>]

Rici Lake wrote:
> ...
> > I did 3 tests. I loaded the strings, got back their char* pointers, 
> > then  put them into the registry and popped them. Then I forced a GC 
> > and loaded the strings in the registry onto the stack and checked 
> > their addresses. They matched. Then I loaded those same strings again 
> > with new pushstring() calls, and they matched. Then I executed these 
> > two functions, in turn:
> >
> > function _getabc()
> >     return "abc"
> > end
> > function _getlongString()
> >     return "AVeryVeryVeryVeryVeryVeryLongString"
> > end
> >
> > Looking at the returned string values, the first one matched and the 
> > second one didn't.
>
> As far as I know, that can only mean that either the long string got 
> gc'd somewhere along the way, or that you mistyped it in the 
> _getlongString function.

OK, thanks for making me go back to look. It was a stupid coding error 
on my part! They are all the same. So that's great. Now I've done a 
timing test between pushstring and rawgeti to push a string. I also 
tested pushlstring() to factor out the strlen() operation. I did 10 
million iterations of pushstring(), pushlstring(), and rawgeti(). The 
results are:

              short string    long string
              ------------    -----------
pushstring()     2.9 sec        4.1 sec
pushlstring()    2.2 sec        3.3 sec
rawgeti()        3.1 sec        3.1 sec

So, hashing short strings that already exist is faster than rawgeti()! 
Since I expect most property strings to be shorter than 35 chars, I 
won't bother with doing rawgeti(). But I will still make an Atom class 
which will cache strings and their lengths and then allow those strings 
to pushed and compared.

I tried hacking this up in my app and saw a 50% improvement in performance!

Anyway, I can post the results if anyone is interested.

Thanks for the help...

--
chris marrin                ,""$,
chris@marrin.com          b`    $                             ,,.
                        mP     b'                            , 1$'
        ,.`           ,b`    ,`                              :$$'
     ,|`             mP    ,`                                       ,mm
   ,b"              b"   ,`            ,mm      m$$    ,m         ,`P$$
  m$`             ,b`  .` ,mm        ,'|$P   ,|"1$`  ,b$P       ,`  :$1
 b$`             ,$: :,`` |$$      ,`   $$` ,|` ,$$,,`"$$     .`    :$|
b$|            _m$`,:`    :$1   ,`     ,$Pm|`    `    :$$,..;"'     |$:
P$b,      _;b$$b$1"       |$$ ,`      ,$$"             ``'          $$
 ```"```'"    `"`         `""`        ""`                          ,P`
"As a general rule,don't solve puzzles that open portals to Hell"'